16 people reported

All time, since Sep 19, including 1 thumbs-up on GitHub.

Unrequested agent fan-outs drain usage

“No cost estimate, budget cap, or confirmation before launch.”

Is this your bug? What gets counted gets fixed. Pile on.

Paste this into your agent. It files your report, with your error and version.

I hit this bug: rapsheet.dev/sheets/claude-code/themes/224/file
Follow it to file my report. Show me the text before you send.

No install. No account. Public and anonymous. The link asks your agent for one report, nothing else. What it sends

Quiet. No reports and thumbs-up in the last two weeks, down from 4 a week at the late-September peak.

5 bugs in this theme

  1. Unrequested multi-agent workflows launch and exhaust usage limits without estimate or confNo workaround yet5 people
  2. Ultracode overspawns agentsNo workaround yet5 people
  3. Subagents spawned despite instructionsNo workaround yet3 people
  4. Subagent fan-out exhausts usage limitNo workaround yet2 people
  5. Review process not scaled to riskNo workaround yet2 people

What's included

  • Model spawns a dozen or more parallel subagents or a Workflow for a small task
  • Runs start with no cost estimate, budget cap or confirmation
  • Standing instructions for light review or no multi-agent work are not followed
  • 5-hour or weekly limit is hit mid-run; agents end and work is lost or repeated
  • Session ends with a bare "You've hit your session limit" message and no handoff

Not included

  • Agents killed at the session limit instead of pausing (separate issue)
  • Subagent cache rewrites after 5-minute TTL or sleep/wake
  • Cloud sessions self-scheduling hourly check-ins that drain credits
  • Long-lived sessions re-reading unbounded context

Original reports16, newest first

  1. Model launches unrequested background subagents despite loaded zero-autonomy rules#104574
  2. Claude Code used up my usage on a simple task with a 36-agent workflow I didn't ask for#103122
  3. Claude launched a 137-agent Workflow for a single-file audit, against a standing "no multi-agent" instruction (~53% of weekly usage in 1 hour)#102248
  4. Severe: Claude Code runs a multi-hour multi-agent pipeline on a UI-only change and ignores explicit ordering instructions#101695
  5. Usage Limit Exhausted in ~20 Seconds with Sonnet 5.5, Max Effort, and Ultracode#101441
  6. Claude Code spawns subagents despite explicit instruction not to#100993
  7. 176904#99775
  8. Excessive token consumption and agent proliferation on simple tasks with Opus model#99759
  9. Orchestrator drifted from the user's ratified light-review tier into fleet-wide full fan-out reviews and burned the weekly usage budget#99376
  10. Parallel subagents consume 642k tokens, then session ends mid-edit without handoff#99125
Show all 16Show only the newest 10
  1. Claude leans towards git worktrees despite includeGitInstructions=false and instructions not to use git#97916
  2. Model launches a dozen parallel subagents and exhausts the account usage limit#97129
  3. You've hit your session limit · resets 2:40am (America/Bogota)#96241
  4. Agent spawning restriction not enforced, causing unauthorized agent fork and excessive API usage#96142
  5. Workflow tool spawns uncapped sub-agents causing excessive token usage and API costs#95816
  6. Agent-initiated multi-agent runs exhaust the 5-hour window with no estimate and no warning#95600

Do you work on Claude Code?

Claim its RAP Sheet to see what your users report and answer them in one place.